In 1796 the Astronomer Royal, Nevil Maskelyne, dismissed his assistant David Kinnebrook. The offence was that Kinnebrook recorded the moment a star crossed the wire of the transit telescope some five to eight tenths of a second later than Maskelyne did. At Greenwich this was not a small thing. The clock depended on the transits, longitude depended on the clock, and the ships depended on longitude. Maskelyne concluded that his assistant had fallen into a vicious way of observing, and let him go.

Twenty-six years later, Bessel at Königsberg read Maskelyne’s published account and did what Maskelyne had not thought to do. He compared his own transit timings against those of astronomers of equal standing, men no one could accuse of a vicious way of observing, and found differences larger than the one that had cost Kinnebrook his post. More importantly he found them steady. Not scatter. A constant personal offset, holding across nights and across stars. He called these involuntary constant differences, and the correction built from them became the personal equation.

Before going any further I want to separate two things that were present that night at the telescope, because the whole of what follows turns on the difference between them.

There was a star crossing a wire, seen. And there was a number, arrived at.

What is not in doubt

What is sensed now is not open to doubt.

Not to doubt, and I have to distinguish that from questioning, because the two get run together. To sense is already a searching. Looking for the wire, waiting for the star, listening to find out whether that was the sound. That is questioning in the mode I mean by the word — open, expanding, prior to any answer — and it is not absent from sensation but constitutive of it.

Doubt is not the opposite of that. Doubt is its perversion. War is the perversion of peace, hate the perversion of love, despair the perversion of sadness — in none of these is the second the absence of the first. The second is made out of the first, with the same energy and the same motion, and what has been removed is the return. Sadness moves and passes; despair is sadness that will not move. Love reaches; hate is love that has fixed. Questioning opens; doubt is questioning that has closed and goes on asking anyway, which is why it cannot arrive and cannot stop.

So doubt requires a claim to be aimed at, and the claim is made afterwards, by the intellect, out of the material. To doubt a sensation is to analyse it, which is an after-the-fact exercise. Of the two, the questioning in the sensing is the furthest thing from doubt there is.

Nor do I mean that the appearance is certain, or sharp, or a foundation on which anything could be built. Sensations are naturally vague. That is the point and not a concession: doubt needs edges to disagree about, and the edges are not sensed. They arrive with the analysis. Nothing survives the moment to be built with anyway.

This is why the child who says the honey is sweet is not making a claim that could be corrected. She is not asserting a theory of sweetness or of honey. The philosopher’s “it appears sweet” is not a correction of her; it is a clarification addressed to grown-ups who have confused themselves.

Everything else is inference. And inference is where error lives, because inference is where the gap is.

Mind is a sense

The felt body appears. Emotion appears. A thought has a texture, an arrival, a duration. These show up in the same way a red patch shows up, and by the criterion above they stand in exactly the same relation to questioning: while occurring, not the sort of thing that can be wrong. The unpleasantness is unpleasant. Whether it is fear, whether it is caused by the letter that came this morning, whether it will still be here at noon — every one of those is inference and every one can fail.

The ordinary objection to taking the inner seriously is that introspection is a bad instrument, and the evidence is real. People confabulate. People report reasons for their behaviour that demonstrably were not the reasons. People fail to notice large changes and then deny that they failed.

I concede all of it, and concede it without reservation, because not one of those findings touches the claim I am making. Confabulation is a false report about a cause. Change blindness is a failed inference about persistence across a gap. Memory revision is an inference from a present appearance to a past one. Each is an error in the layer where errors are possible. None is a case of something appearing wrongly while it appeared.

So the psychology of introspective failure is not an argument against the immediate. It is a map of the inferential layer, made with care, and I would rather have it than not.

The induction of inductions

Now go back to Königsberg.

What Bessel had of the star was a crossing, seen, now. What he had of the offset was something else entirely: a comparison of records made on different nights by different men, a judgment that the differences were systematic rather than random, and a conclusion that what had held would go on holding.

Hume named the difficulty and I will not spend long on it. The inference requires that unobserved cases resemble observed ones, and that principle cannot be established deductively without assuming itself. He concluded that induction is not the work of reason but of the imagination.

The general form is what I am after, and it is this. All replicated measurements, never finished, may be treated as finished.

That is the induction of inductions, and it is the same move as the axiom of completeness, made outside the language of numbers. The structure matches at every point. A sequence of replications, each refining the estimate. Bounded, since the value is not infinite. Convergent. And never finished, because the experiment can always be run once more, and there is always a further place of precision the method has not reached. Then the move: assert that the sequence has a limit, and call the limit the true value.

Consensus is the least upper bound.

It is two-storied, which is why it is an induction upon inductions and not simply an induction. Each single measurement is already one — that the instrument reads true, that conditions held while nobody was watching. Replication then inducts across those inductions and takes their limit. The second storey is where the metaphysical work is done, because it is there that the observers are supposed to drop out. At the limit, so the story goes, the personal offsets cancel and what remains is the object itself, unmediated. That is the escape from the first person. It is asserted, not found.

The error bar is the “…”. It is notation standing in for the rest of an investigation that nobody carried out. A meta-analysis is a sequence of studies with a limit taken.

And the cost is the cost I described for the plank. The questions live at each level of precision — is that established enough, has that been replicated enough, what would the next study show — and they are answered by declaring the unfinished sequence finished. No room is left in the established value for doubt to stand, which is the one place doubt actually belongs, since it is the inferential layer and nowhere else.

But notice what else gets shut out, and why. The motive for taking the limit is not curiosity. It is the craving for certainty, which is the same thing as the fear of doubt, and the two names describe one condition. A mathematician cannot rest in an unfinished sequence. A profession cannot rest in a result that might not replicate. So the sequence is closed by decree — and because doubt and questioning are made of the same material, closing the door against the one locks out the other. What is expelled is not only the corrosive asking that could not stop. It is the open asking that would have gone on opening. The axiom does not distinguish between them, and cannot, because they arrive through the same door.

This is also why the enterprise heats the mind rather than cooling it. Craving is the heat. A discipline organized around wanting an end to uncertainty, and around dreading its return, will generate exactly the temperature it is trying to measure away — but that is a longer argument and belongs elsewhere.

There is a second route to the same finding, which does not pass through Hume or through mathematics. A person who sits still and watches something he had taken to be steady will see it go. He does not deduce impermanence; he observes it, repeatedly, and the observation is of the kind described above — not open to doubt while it occurs. That two methods this different arrive at the same place is worth more than either arriving alone. It is also the thesis of this essay demonstrated rather than argued: a finding of trained attention and a finding of philosophical analysis converging on the status of constants.

Comparing observers requires an observer

Here is the point at which the standard picture cannot be repaired.

Bessel did not step outside himself to discover that he and his colleagues differed. He looked at the records. He judged that the difference was systematic and not noise. He decided how many nights were enough. Every one of those acts is precisely the kind of act the personal equation exists to correct for, and every one was performed in the first person, unassisted, by a man with an offset of his own.

There is no regress-stopper. There never has been. Each correction of first-person judgment is designed, applied and evaluated by first-person judgment, and if you go looking for the level at which this stops being true you will not find it. You will only find people, looking, deciding.

Which means the apparatus of objectivity is not a way of getting out. It is a way of getting a better result from inside, and the difference matters enormously in practice and not at all in principle.

Independence, and the machinery that spends it

The serious defence of third-person method is not that any individual scientist errs less than any contemplative. It is that errors are uncorrelated. My misreading and yours are independent, so when we pool, noise cancels and signal stands.

So look at what the profession does to produce its agreements. Shared training. Common instruments from common manufacturers. A standardized vocabulary in which the phenomenon may be described. Peer review by people trained in that vocabulary. Replication by protocols fixed in advance. Textbooks that settle what a competent observation is before anyone observes anything.

Every one of these is a correlating operation. Each is built, deliberately and well, to make two observers more alike. Which means the agreement they produce cannot then be cited as independent convergence upon the object, because independence was the thing being removed. Consensus is both the output of the correction machinery and an input to it, and the relation runs the wrong way for comfort: the tighter the correction, the less the eventual agreement tells you.

I am not saying the corrections are bad. They are why the bridges stand. I am saying they purchase reliability by spending independence, and you cannot spend it and still have it.

Put beside the previous section, the two halves lock. The correlating machinery is what makes the sequence converge. The convergence is then treated as arrival at the object. So the sequence is engineered to have a limit, and the limit is cited as evidence about what it converged upon. This is also why the sociological critique of guilds and the epistemological defence of method are not two conversations. They are one. The institution erodes its own justification in the act of performing it.

What survives, and why it costs

Something survives, and on the spine of this essay it comes clear.

Some observations arrive unbidden, contradict what the observer wanted, and cannot be argued away. The bridge falls. The trial fails. The result will not replicate no matter who runs it. That kind of observation is worth a great deal.

I used to think this was a second and separate virtue, and wondered whether it was only independence wearing a coat. I no longer think so. What is happening in those cases is that an appearance has overruled an inference. The un-questionable has met the questionable, and won. That is why it costs: the whole inferential structure had money on the other outcome, and the appearance does not negotiate.

And this can happen in either domain. It happens to the engineer standing in front of a structure that is not behaving. It happens to the practitioner who sees, plainly and unwillingly, what cuts against his teacher, his doctrine and his hopes for himself. Same event. Same currency.

The inversion

The standard picture has trained attention waiting outside, hoping to be licensed by objective method — hoping that one day the disciplines of the inner will be validated by instruments and thereby allowed to count.

The dependency runs the other way.

Discriminating what is present from what you expected to be present. Noticing that you are tired, that you are invested, that you have already decided. Holding attention steady enough to register a small change nothing prompted you to look for. Every gauge-reading in the history of the world has run on these.

And when you examine what the apparatus consists of, you find prosthetics for known defects in exactly these capacities. Blinding manages expectation, and expectation is an inner phenomenon. Pre-registration manages the drift of a hope into a hypothesis. Instrumentation replaces a judgment that fatigues with a trace that does not. The personal equation corrects a constant in a man’s attention that the man could not feel.

You cannot build a prosthetic without knowing what is failing. Each device encodes a piece of first-person knowledge — a fact about attention, discovered by attention — and then conceals its origin by working so well that no one needs to think about the defect again.

Attention is not the worse instrument. It is the instrument the others are corrections to.

The line and the circle

There are two shapes a correction can take, and the difference is the difference between an equation and a return.

Identity is a line. Even written as A = A it requires two positions and a sign between them holding the gap closed; you cannot write “A =” and have said anything. Selfhood on that model is an assertion of sameness across a distance, and it needs the gap covered, exactly as the axiom covers the gaps in the line.

The other shape needs no equals sign. One object, and coming back to it. Aristotle’s point that circular motion cannot be defined by an opposition is the relevant one: the present you, reading, and returning to the present you, reading, is enough for motion, and there is no second term. Nothing is claimed to be the same. Heraclitus has it already — we step and do not step in the same rivers.

Meditation is the second shape, and this makes it a calibration of a different kind from Bessel’s. Both involve return. But a linear correction has an output: a number you subtract, after which the matter need not be attended to again. The return produces no number. Nothing is stored, nothing accrues, and tomorrow you do it again with no credit from today. The output of the practice is the practice.

That is why the objectivity apparatus can forget its origin and the practice cannot.

And the two treat wandering in opposite ways. Blinding, protocol, pre-registration, fixed vocabulary — every one is engineered to prevent the mind from going where it will. The return does not prevent it. It consents to it: an awareness and a deliberateness about where one goes with one’s mind, rather than a wall against the going. One approach removes the wandering. The other keeps the wandering and adds the coming back.

Which gives the perversion its shape. Doubt is wandering with the return taken out, and an apparatus built of protocols against a wandering never permitted to come back does not contain doubt. It institutionalizes it, and then covers the hole with a limit so that the asking stops by decree rather than by arrival.

Constants, and how long we have been looking

The impermanence of constants is not a mystical objection and I want to state it carefully, because the careless version is easy to dismiss.

Physics does check this. Variation in the fine-structure constant has been probed against the natural fission reactor at Oklo, two billion years old, against absorption lines in quasar spectra, against the abundances of the light elements formed in the first minutes. The bounds that come back are tight, and anyone who says nobody has looked deserves to be stopped at the first paragraph.

But every one of those bounds is derived inside a framework that assumes the stability it is testing. The Oklo constraint is read out of nuclear physics taken to have been the same nuclear physics. This is not a scandal and there is no way around it. It is what testing a constant necessarily looks like from inside.

What can be said plainly is the ratio. We have been measuring carefully for something on the order of two hundred years. We license conclusions across thirteen point eight billion — a figure itself computed on the assumption that the relevant constants held throughout. Two hundred against thirteen point eight billion is not a small extrapolation shored up by good behaviour in the sample. It is an act of faith with an error bar printed on it, and the error bar is calculated using the assumption in question.

The constant feels constant. That feeling is a fervour about something only time can report on, and we have not had the time.

Two things I am not saying

That a profession’s first job is self-preservation, I hold to. Its first job, not its only one. People inside these institutions do careful, honest and expensive work, and much of it survives the pressure it is under. If “first” slides into “only,” the argument becomes a complaint, and a complaint is easy to put down.

And contemplative institutions gatekeep at least as hard as universities. Lineage, authorization, the approved description of what you saw, the quiet disappearance of the student whose experience did not fit the map. I concede this early because it costs the argument nothing. My claim concerns the structure of investigation, not which guild has nicer people. Every guild correlates its members. That is what a guild is for.

Where this sits

Husserl described objective science as a construction out of the lifeworld that forgot it was a construction. Merleau-Ponty treated the objective attitude as a derived posture concealing its origin. James took introspection as a primary method and gave it the seriousness of any laboratory procedure. Fleck showed that a thought-collective determines what can be seen at all. Gieryn named the boundary work by which a profession draws the line around itself. Daston and Galison traced mechanical objectivity as a discipline of self-restraint — the observer suppressing his own judgment, an inner exercise throughout. Varela built a research programme on the dependency instead of arguing about it.

What I have not found stated cleanly is that the tools of objectivity are made out of first-person knowledge and cannot be made any other way, which reverses the licensing relation. Nor have I found anyone using Bessel for it, though he is sitting right there.

Back to Königsberg

There is a detail I have saved.

Having established the personal equation, Bessel found that it drifts. The constant is not constant. An observer’s offset changes with the year, with the instrument, with the man. So the correction must be remade, and remaking it means going back to the same place: a person attending to a star and to himself at once, taking the measure of his own attending.

I want to mark the next sentence as mine rather than his. Bessel found his constant by comparing observers, not by any observer returning to himself, so what I am about to say is an interpretation laid over the history and not something the history states. But the drift is the practice showing through the description. A return was being modelled as a line, and the line would not hold, because each re-measurement is a coming back to the same observer at the same instrument, claiming nothing about sameness and getting contact instead. Astronomy got the practice right and the description wrong, and it was the description that was institutionalized.

We have been going back ever since. We only stopped saying so.

And there is a reason to care beyond the accounting. If the immediate is not in doubt, then the endless refinement toward a precision that never arrives is the intellect sharpening a vagueness it did not find sharp, and the appearances we were told to look past were never deficient. What follows from that is happiness, which is a longer argument and one I have made elsewhere. Here I only want the ground it stands on: this, now, sensed, before the question can find its footing.